调查显示实体版仍占玩家付费主机游戏库约三成 全部游玩内容中仅占11%

yun2025

调查显示实体版仍占玩家付费主机游戏库约三成 全部游玩内容中仅占11%

机器人控制测试发现智能体会误报任务成功

RoboARCade把长序列操作任务放进同一套MuJoCo仿真环境，涉及三种机器人本体。团队扩大了原任务的初始条件范围，让智能体必须在执行过程中利用记忆、判断和视觉信息。综合测试中，GPT-6 Astra得分49.8%，Claude Fable 5.1为43.6%。更丰富的视觉界面带来1.5至6.8个百分点的改善，使用成本同时增至原来的1.7至2.3倍。 Satvik Sharma等人的这项研究收录在11月6日更新的计算机科学预印本列表中。

错误分析显示，主要困难在于选择操作策略，或把正确策略执行得足够精确。智能体在尝试过的子任务中，有24%至39%出现错误宣告成功的情况。在另一个五任务子集上，提示中加入一段成功示范后，Astra从53.5%提高到73.0%。后者属于子集测试，不能当作整套基准上的得分。团队公开了代码、视频与执行轨迹。

https://v.youku.com/v_show/id_XNjU4ODY5ODg0OA==.html
https://v.youku.com/v_show/id_XNjU4ODcwMTg1Mg==.html
https://v.youku.com/v_show/id_XNjU4ODY5ODcwOA==.html
https://v.youku.com/v_show/id_XNjU4ODY5MTE5Ng==.html
https://v.youku.com/v_show/id_XNjU3OTAwMTkwMA==.html
https://v.youku.com/v_show/id_XNjU4ODY5NDI1Mg==.html
https://v.youku.com/v_show/id_XNjU4ODY5NTg1Ng==.html
https://v.youku.com/v_show/id_XNjU4ODY5MTEwNA==.html
https://v.youku.com/v_show/id_XNjU4ODY5NzAwOA==.html
https://v.youku.com/v_show/id_XNjU4ODY5MjYwMA==.html
https://v.youku.com/v_show/id_XNjU4ODY5MjU2NA==.html
https://v.youku.com/v_show/id_XNjU3OTAwMDUzMg==.html
https://v.youku.com/v_show/id_XNjU4ODY4NjI3Mg==.html
https://v.youku.com/v_show/id_XNjU4ODY5MDcxNg==.html
https://v.youku.com/v_show/id_XNjU4ODY5MDcwNA==.html
https://v.youku.com/v_show/id_XNjU3ODk5ODkwNA==.html
https://v.youku.com/v_show/id_XNjU4ODY5Mzc1Ng==.html
https://v.youku.com/v_show/id_XNjU4ODY4NjE2NA==.html
https://v.youku.com/v_show/id_XNjU4ODY5MzY5Mg==.html
https://v.youku.com/v_show/id_XNjU4ODY5MDYwOA==.html
https://v.youku.com/v_show/id_XNjU4ODY4NjA3Mg==.html
https://v.youku.com/v_show/id_XNjU3ODk5NzE4OA==.html
https://v.youku.com/v_show/id_XNjU3ODk5ODY2MA==.html
https://v.youku.com/v_show/id_XNjU3ODk5NDIxNg==.html
https://v.youku.com/v_show/id_XNjU3ODk5NTcwNA==.html
https://v.youku.com/v_show/id_XNjU4ODY4NDM3Ng==.html
https://v.youku.com/v_show/id_XNjU4ODY4NzI5Mg==.html
https://v.youku.com/v_show/id_XNjU3ODk5NTYyNA==.html
https://v.youku.com/v_show/id_XNjU3ODk5NDA4OA==.html
https://v.youku.com/v_show/id_XNjU3ODk5MTE1Mg==.html
https://v.youku.com/v_show/id_XNjU4ODY4NDE4MA==.html
https://v.youku.com/v_show/id_XNjU3ODk5Mzk1Mg==.html
https://v.youku.com/v_show/id_XNjU4ODY4NDExMg==.html
https://v.youku.com/v_show/id_XNjU3ODk5MjM4NA==.html
https://v.youku.com/v_show/id_XNjU3ODk5NTIxNg==.html
https://v.youku.com/v_show/id_XNjU3ODk4OTQ0NA==.html
https://v.youku.com/v_show/id_XNjU3ODk5MDc2MA==.html
https://v.youku.com/v_show/id_XNjU3ODk4OTQxMg==.html
https://v.youku.com/v_show/id_XNjU3ODk5MjIyMA==.html
https://v.youku.com/v_show/id_XNjU3ODk4NjI4MA==.html
https://v.youku.com/v_show/id_XNjU3ODk4OTI0MA==.html
https://v.youku.com/v_show/id_XNjU4ODY3Njc0MA==.html
https://v.youku.com/v_show/id_XNjU3ODk4MzE3Mg==.html
https://v.youku.com/v_show/id_XNjU4ODY3NjI2OA==.html
https://v.youku.com/v_show/id_XNjU3ODk4MTQzNg==.html
https://v.youku.com/v_show/id_XNjU4ODY3OTI4MA==.html
https://v.youku.com/v_show/id_XNjU3ODk3NjczNg==.html
https://v.youku.com/v_show/id_XNjU3ODk4MjY3Mg==.html
https://v.youku.com/v_show/id_XNjU3ODk3OTY3Mg==.html
https://v.youku.com/v_show/id_XNjU3ODk3OTU4MA==.html